Paper Info
Title
Understanding and Improving Sequence-to-Sequence Pretraining for Neural Machine Translation
Abstract
In this paper, we present a substantial step in better understanding the SOTA sequence-to-sequence (Seq2Seq) pretraining for neural machine translation (NMT). We focus on studying the impact of the jointly pretrained decoder, which is the main difference between Seq2Seq pretraining and previous encoderbased pretraining approaches for NMT. By carefully designing experiments on three language pairs, we find that Seq2Seq pretraining is a double-edged sword: On one hand, it helps NMT models to produce more diverse translations and reduce adequacy-related translation errors. On the other hand, the discrepancies between Seq2Seq pretraining and NMT finetuning limit the translation quality (i.e., domain discrepancy) and induce the over-estimation issue (i.e., objective discrepancy). Based on these observations, we further propose simple and effective strategies, named in-domain pretraining and input adaptation to remedy the domain and objective discrepancies, respectively. Experimental results on several language pairs show that our approach can consistently improve both translation performance and model robustness upon Seq2Seq pretraining.
Year
DOI
Venue
2022
10.18653/v1/2022.acl-long.185
PROCEEDINGS OF THE 60TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (ACL 2022), VOL 1: (LONG PAPERS)
DocType
Volume
Citations 
Conference
Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
0
PageRank 
References 
Authors
0.34
0
7
Name
Order
Citations
PageRank
Wenxuan Wang100.34
Wenxiang Jiao222.39
Yongchang Hao300.34
Xing Wang45810.07
Shuming Shi562058.27
Zhaopeng Tu651839.95
Michael R. Lyu763.15